geekjock

It seems like companies that traditionally offered multi-tenant cloud and on-premises versions of their software are now switching to offering “private cloud” or “managed service” versions that are a hybrid of the previous approaches.

From what I can tell, this approach involves the vendor creating and managing a customer’s software instance, but doing so in a single-tenant model on the customer’s preferred cloud provider.

I’m curious to know how this works. Does the vendor get access the customer’s AWS/Azure/GC account and then create the resources to run the software there? Or does the vendor deploy the software in their own cloud accounts in the region/provider preferred by the customer, and then somehow keep the instance confined to being accessed within the customer’s network?

Example: https://document360.com/private-hosting/

I was looking at the GitHub GraphQL API (https://developer.github.com/v4/query) and noticed that they only expose a few root query types.

For example, to get a pull request you must use the `repository` Query type and access a pull request through the `Repository` object fields.

  {
    repository(name: 'test') {
      pull_request(number: 2) {
        title
      }
    }
  }

You can't just access a pull request by ID via a root query type:

  {
    pull_request(id: 12345) {
      title
    }
  }

Is this a recommended practice? Is there an argument for limiting root query types in favor of using nested object types to keep the schema cleaner?